Follicular data

Introduction

Outline

• Experimental substrates
• QC stats
• Preliminary analysis
• Next steps

Experimental substrates

Data for 5 pre- and post-treatment pairs:

Transformed FL:

• FL1004
• FL1012
• FL1018

Progressed FL:

• FL2001

Data

class: SingleCellExperiment 
dim: 33694 28063 
metadata(0):
assays(1): counts
rownames(33694): ENSG00000243485 ENSG00000237613 ... ENSG00000277475 ENSG00000268674
rowData names(2): ID Symbol
colnames: NULL
colData names(10): Sample Barcode ... status batch
reducedDimNames(0):
spikeNames(0):

Mitochondrial counts

`legend.margin` must be specified using `margin()`. For the old behavior use legend.spacing

Ribosomal counts

`legend.margin` must be specified using `margin()`. For the old behavior use legend.spacing

Filtering

Normalization

Total number of UMIs

Cell cycle

Batch correction (for visualization)

Mutual nearest neighbours

Mutual nearest neighbours is a batch correction method that finds nearest neighbours across batches to correct other cells by.

We don’t have spikes to use here, so we’re making an assumption that most noise is technical even for endogenous genes.

CellAssign

Conservative

WARNING:tensorflow:From /home/rstudio/miniconda/envs/r-tensorflow/lib/python3.6/site-packages/tensorflow/python/util/deprecation.py:302: NegativeBinomial.__init__ (from tensorflow.contrib.distributions.python.ops.negative_binomial) is deprecated and will be removed after 2018-10-01.
Instructions for updating:
The TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use `tfp.distributions` instead of `tf.contrib.distributions`.
2018-10-26 14:38:27.978413: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
440 L old: -1727070.14330226; L new: -324916.155194938; Difference (%): 0.811868581913136
80  L old: -324916.155194938; L new: -322873.153686804; Difference (%): 0.0062877806334611
40  L old: -322873.153686804; L new: -322794.15493522; Difference (%): 0.000244674265055047
20  L old: -322794.15493522; L new: -322771.418983687; Difference (%): 7.0434830323801e-05
20  L old: -322771.418983687; L new: -322758.562220461; Difference (%): 3.98324091617553e-05
20  L old: -322758.562220461; L new: -322748.893090675; Difference (%): 2.9957779334512e-05
20  L old: -322748.893090675; L new: -322740.823090444; Difference (%): 2.50039594392631e-05
20  L old: -322740.823090444; L new: -322734.018702476; Difference (%): 2.10831338369367e-05
20  L old: -322734.018702476; L new: -322728.250758228; Difference (%): 1.78721297197567e-05
20  L old: -322728.250758228; L new: -322723.338487699; Difference (%): 1.52210738224163e-05
20  L old: -322723.338487699; L new: -322719.138389789; Difference (%): 1.30145465447879e-05
20  L old: -322719.138389789; L new: -322715.534754447; Difference (%): 1.11664754688296e-05
20  L old: -322715.534754447; L new: -322712.4332455; Difference (%): 9.61065896599896e-06

T-cell subclustering

Full

WARNING:tensorflow:From /home/rstudio/miniconda/envs/r-tensorflow/lib/python3.6/site-packages/tensorflow/python/util/deprecation.py:302: NegativeBinomial.__init__ (from tensorflow.contrib.distributions.python.ops.negative_binomial) is deprecated and will be removed after 2018-10-01.
Instructions for updating:
The TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use `tfp.distributions` instead of `tf.contrib.distributions`.
2018-10-25 19:35:22.661101: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
440 L old: -1460113.06772068; L new: -429822.150100088; Difference (%): 0.705624064599966
160 L old: -429822.150100088; L new: -417067.097140167; Difference (%): 0.0296751876490095
60  L old: -417067.097140167; L new: -416106.956773512; Difference (%): 0.00230212446207986
60  L old: -416106.956773512; L new: -415725.356357592; Difference (%): 0.000917072905675379
40  L old: -415725.356357592; L new: -415568.054457712; Difference (%): 0.000378379373482875
20  L old: -415568.054457712; L new: -415517.410467035; Difference (%): 0.000121866900340133
20  L old: -415517.410467035; L new: -415480.453148442; Difference (%): 8.8942888222262e-05
20  L old: -415480.453148442; L new: -415454.431146079; Difference (%): 6.26311109593666e-05
20  L old: -415454.431146079; L new: -415435.115038928; Difference (%): 4.64939249722595e-05
20  L old: -415435.115038928; L new: -415419.943479967; Difference (%): 3.65196836092371e-05
20  L old: -415419.943479967; L new: -415408.110328068; Difference (%): 2.8484794927761e-05
20  L old: -415408.110328068; L new: -415398.85243704; Difference (%): 2.22862548830485e-05
20  L old: -415398.85243704; L new: -415391.441770564; Difference (%): 1.78398819181487e-05
20  L old: -415391.441770564; L new: -415385.38857749; Difference (%): 1.45722623668343e-05
20  L old: -415385.38857749; L new: -415380.290312505; Difference (%): 1.22735780439428e-05
20  L old: -415380.290312505; L new: -415376.146127331; Difference (%): 9.97684596495281e-06

Probabilities

Match to unsupervised clusters

Malignant vs. nonmalignant B cell identification

Patient-by-patient analysis

FL1004

Finding 30 nearest neighbors using minkowski metric and 'auto' algorithm
Neighbors computed in 0.10490250587463379 seconds
Jaccard graph constructed in 0.7177457809448242 seconds
Wrote graph to binary file in 0.10836005210876465 seconds
Running Louvain modularity optimization
After 1 runs, maximum modularity is Q = 0.688699
Louvain completed 21 runs in 0.6867151260375977 seconds
PhenoGraph complete in 1.6278762817382812 seconds

FL1012

FL1018

FL2001

Joining, by = \sample_barcode\

Proportions and differential expression

rr pathway_timepoint_cytotoxic <- gage_analysis(sce = sce_follicular_normalized_labeled_full %>% scater::filter(celltype == T cells), comparison_type = , filter_ribo = filter_ribo, filter_mito = filter_mito, gene_set_path = /datadrive/projects/cellassign-analysis/downloads/c2.cp.reactome.v6.2.symbols.gmt, rowdat = rowData(sce_follicular_normalized_labeled_full))

plot_markers(pathway_timepoint_cytotoxic$pathways %>% plyr::rename(c(‘pathway’=‘Symbol’, ‘q.val’=‘FDR’, ‘mean_logFC’=‘logFC’)) %>% dplyr::mutate(is_significant = FDR <= 0.05), top_n = 5, log_q_val_threshold = NULL, nrow = 1, subtitle = 2 vs. T1, top_direction = )

Upregulation of PD-1 (PD-1 blockade?) in cytotoxic T cells at T2? CHECK if this corresponds to membership in the IFNG-containing subcluster of CD8 T cells.

rr pathway_timepoint_naive <- gage_analysis(sce = sce_follicular_normalized_labeled_full %>% scater::filter(celltype == /CM CD4), comparison_type = , filter_ribo = filter_ribo, filter_mito = filter_mito, gene_set_path = /datadrive/projects/cellassign-analysis/downloads/c2.cp.reactome.v6.2.symbols.gmt, rowdat = rowData(sce_follicular_normalized_labeled_full))

plot_markers(pathway_timepoint_naive$pathways %>% plyr::rename(c(‘pathway’=‘Symbol’, ‘q.val’=‘FDR’, ‘mean_logFC’=‘logFC’)) %>% dplyr::mutate(is_significant = FDR <= 0.05), top_n = 5, log_q_val_threshold = NULL, nrow = 1, subtitle = 2 vs. T1, top_direction = )

rr pathway_timepoint_naive$pathways %>% dplyr::filter(q.val < 0.05) %>% dplyr::arrange(-mean_logFC) %>% head(10)

rr pathway_timepoint_malignantb <- gage_analysis(sce = sce_follicular_normalized_labeled_full %>% scater::filter(celltype == cells & malignant_status_manual == ), comparison_type = , filter_ribo = filter_ribo, filter_mito = filter_mito, gene_set_path = /datadrive/projects/cellassign-analysis/downloads/c2.cp.reactome.v6.2.symbols.gmt, rowdat = rowData(sce_follicular_normalized_labeled_full))

Joining by: pathway

rr plot_markers(pathway_timepoint_malignantb$pathways %>% plyr::rename(c(‘pathway’=‘Symbol’, ‘q.val’=‘FDR’, ‘mean_logFC’=‘logFC’)) %>% dplyr::mutate(is_significant = FDR <= 0.05), top_n = 5, log_q_val_threshold = NULL, nrow = 1, subtitle = 2 vs. T1, top_direction = )

rr pathway_timepoint_malignantb$pathways %>% dplyr::filter(q.val < 0.05) %>% dplyr::arrange(-mean_logFC) %>% head(10)

rr pathway_timepoint_nonmalignantb <- gage_analysis(sce = sce_follicular_normalized_labeled_full %>% scater::filter(celltype == cells & malignant_status_manual == ), comparison_type = , filter_ribo = filter_ribo, filter_mito = filter_mito, gene_set_path = /datadrive/projects/cellassign-analysis/downloads/c2.cp.reactome.v6.2.symbols.gmt, rowdat = rowData(sce_follicular_normalized_labeled_full))

plot_markers(pathway_timepoint_nonmalignantb$pathways %>% plyr::rename(c(‘pathway’=‘Symbol’, ‘q.val’=‘FDR’, ‘mean_logFC’=‘logFC’)) %>% dplyr::mutate(is_significant = FDR <= 0.05), top_n = 5, log_q_val_threshold = NULL, nrow = 1, subtitle = 2 vs. T1, top_direction = )

rr pathway_timepoint_FL1018malignantb <- gage_analysis(sce = sce_follicular_normalized_labeled_full %>% scater::filter(celltype == cells & malignant_status_manual ==  & patient == 1018), comparison_type = , filter_ribo = filter_ribo, filter_mito = filter_mito, gene_set_path = /datadrive/projects/cellassign-analysis/downloads/c2.cp.reactome.v6.2.symbols.gmt, rowdat = rowData(sce_follicular_normalized_labeled_full))

Joining by: pathway

FACS validation

Should be run with minimally-filtered dataset (i.e. filtered at 10% mito or higher), as opposed to a 5% filter that could be used for analysis.

Data

T cell proportion

CD4 T cell (out of T cell) proportion

Tfh out of CD4 T cell proportion

Generally fairly consistent. Note that quantifying memory T cells is a problem since we don’t have a good marker to distinguish naive from CM.